A Low Cost Desktop Robot and Tele-Presence Device for Interactive Speech Research

Michael C. BradyInterspeech 2016

TL;DR

This paper introduces a low-cost, open-source, and modular robotic platform designed to bridge the gap between speech engineers and robotics engineers. By providing an accessible hardware and software framework, the authors aim to foster collaboration and accelerate research in human-robot speech interaction.

What Problem Does This Paper Address?

Developing effective human-robot speech interaction systems is hindered by a "diffusion of responsibility." Speech engineers often lack access to the expensive robotic hardware needed for development, while robotics engineers may not have the specialized speech expertise. This disconnect slows progress in creating robots that can truly understand and respond to human speech in a nuanced way, especially when considering the need for robots to reason about commands and their consequences. The authors aim to remedy this by creating a standardized, affordable, and customizable platform that encourages joint development.

Methods

The core of the paper is the introduction of a low-cost, interactive embodied speech device. The platform is constructed using readily available off-the-shelf components, 3D-printed parts, and laser-cut elements. The authors provide freely available design files for these custom parts, along with assembly guides and component shopping lists.

Two versions of the system are presented:

  • Minimal System: A basic desktop robot focused on speech interaction.
  • Elaborate System: An enhanced version that includes robotic arms and an Xbox One Kinect V2 sensor for more complex interaction capabilities.

For software, the authors offer:

  • A basic software framework in both MATLAB and C/C++ for direct motor control, bypassing the complexity of full ROS for initial development.
  • Template code for a ROS node to interface with the device, catering to those already familiar with the Robot Operating System.
  • An open-source project and user forum to facilitate community collaboration and the sharing of developed components.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Democratizing Robotics for Speech Research: The most significant insight is the creation of an affordable and accessible platform that lowers the barrier to entry for researchers interested in human-robot speech interaction. This directly addresses the "diffusion of responsibility" problem.
  • Modularity and Customization: The design emphasizes modularity, allowing users to build a basic system or expand it with features like robotic arms and advanced sensors, catering to diverse research needs.
  • Open-Source Ecosystem: The commitment to providing design files, software frameworks, and a community forum is crucial for fostering a collaborative environment and building a shared knowledge base.
  • Bridging Disciplines: The platform is designed to be approachable for both speech engineers and robotics engineers, encouraging interdisciplinary work by offering tools and frameworks familiar to each group (e.g., MATLAB for speech/signal processing, ROS for robotics).

Key Results

The paper primarily presents a system design and a framework for research, rather than quantitative experimental results. The key "results" are the availability of:

  • A low-cost robotic platform (parts cost less than a laptop).
  • Freely available design files for 3D-printed and laser-cut components.
  • Basic software frameworks for motor control (MATLAB, C/C++) and ROS integration.
  • A community forum for discussion and development.

The elaborate system with two arms and a Kinect weighs approximately 13 pounds and can be desktop-mounted or placed on a microphone stand.

Limitations and Future Work

Limitations: * The paper focuses on the hardware and software framework itself, with limited discussion of specific interactive speech applications or evaluations of the system's performance in real-world scenarios. * While MATLAB is highlighted for its accessibility, the reliance on commercial software might still be a barrier for some. The C/C++ and ROS options aim to mitigate this.

Future Work: * The authors eagerly seek community feedback to guide the development of the platform and its associated software. * They anticipate that progress will build upon existing standards in conversational avatars (e.g., Perception Markup Language, Behavior Markup Language) and robotics (e.g., ROS). * The paper poses an "open and exciting question" about how developers will ultimately use this platform and what new directions speech software development will take.

Summary updated

BibTeX
@inproceedings{brady2016low,
  author = {Michael C. Brady},
  title = {{A Low Cost Desktop Robot and Tele-Presence Device for Interactive Speech Research}},
  booktitle = {Interspeech},
  year = {2016},
}
Figure